-------------------------------------------------------------------------------
help for trimplot
-------------------------------------------------------------------------------

Plots of trimmed means 

        trimplot varname [if exp] [in range] [, over(varname) percent
                 scatter_options]

        trimplot varlist [if exp] [in range] [, percent scatter_options]


Description

    trimplot produces plots of trimmed means versus depth for one or more
    numeric variables. Such plots may help specifically in choosing or
    assessing measures of level and generally in assessing the symmetry or
    skewness of distributions. They can be used to compare distributions or
    to assess whether transformations are necessary or effective.

    trimplot may be used to show trimmed means for one variable, in which
    case different groups may be distinguished by the over() option; or for
    several variables.


Remarks 

    Order n data values for a variable x and label them such that x(1) <= ...
    <= x(n). Following Tukey (1977), depth is defined as 1 for x(1) and x(n),
    2 for x(2) and x(n-1), and so forth: it is the smaller number reached by
    counting inwards from either extreme x(1) or x(n) toward any specified
    value. So the depth of x(i) is the smaller of i and n - i + 1.

    Trimmed means may be related to depth as follows. A trimmed mean may be
    defined for any particular depth as the mean of all values with that
    depth or greater. Thus the trimmed mean for depth 1 is the mean of all
    values. The trimmed mean for depth 2 is the mean of all values except
    those of depth 1, i.e. all values except for the extremes. The trimmed
    mean for depth 3 is the mean of all values except those of depth 1 and 2;
    and so forth.

    The highest depth observed for a distribution occurs once if n is odd and
    twice if n is even; either way it labels those values whose mean is the
    median. Thus trimmed means range from the mean to the median.

    The idea of plotting trimmed mean versus percent trimmed can only be a
    little deal. An example can be found in Rosenberger and Gasko (1983,
    p.315). Users knowing good and/or early references are welcome to email
    me with details.

    For more on trimmed means, see the help for trimmean (which must be
    installed first).


Options 

    over(varname) specifies that calculations are to be carried out
        separately for each group defined by varname.  over() is allowed only
        with a single variable to be plotted.

    percent specifies that depth is to be scaled and plotted as percent
        trimmed, which will range from 0 to nearly 50 (a median cannot be
        based on no observed values, so 50 cannot be attained).

    scatter_options are options of twoway scatter.


Examples

    . webuse citytemp
    . describe
    . trimplot *dd
    . trimplot temp*
    . trimplot tempjan, over(region) percent


Author

    Nicholas J. Cox, Durham University
    n.j.cox@durham.ac.uk


Acknowledgments 

    Ariel Linden found a typo in the help.


References

    Rosenberger, J.L. and Gasko, M. 1983.  Comparing location estimators:
        trimmed means, medians, and trimean.  In Hoaglin, D.C., Mosteller, F.
        and Tukey, J.W. (Eds) Understanding robust and exploratory data
        analysis.  New York: John Wiley, 297-338.

    Tukey, J.W. 1977.  Exploratory data analysis.  Reading, MA:
        Addison-Wesley.


Also see 

    summarize, means, trimmean (if installed), hsmode (if installed), shorth
             (if installed)